A digital psychological behavior tic monitoring method and system

Abstract

The application provides a digital psychological behavior tic monitoring method and system, relates to the technical field of digital psychological behavior tic monitoring, establishes a tic monitoring coordinate system, determines an initial static node, acquires node dynamic behavior track information of the initial static node, deletes origin displacement information, and acquires displacement behavior track information; difference anomaly analysis and soft clustering anomaly analysis are performed on the initial static node through an AI chip, and then anomaly confidence analysis is performed; corresponding anomaly nodes are acquired according to the obtained difference confidence analysis result and soft clustering confidence analysis result, tic confidence determination is performed, tic confidence determination information is acquired, and the AI chip can realize digital anomaly analysis on human psychological behavior tics, so that the analysis efficiency and accuracy are greatly improved.

Description

Technical Field

[0001] This invention proposes a digital psychological and behavioral tic monitoring method and system, which relates to the field of tic monitoring technology, specifically to the field of digital psychological and behavioral tic monitoring technology. Background Technology

[0002] Tourette syndrome is a common neurodevelopmental disorder. Traditional tic monitoring relies on direct observation by clinicians or patient self-reporting, which suffers from high subjectivity, limited monitoring time and location, and difficulty in capturing occasional tic behaviors. Furthermore, it cannot achieve remote, long-term dynamic monitoring using AI chips. Existing technologies mostly focus on single-dimensional feature extraction, making them susceptible to interference factors, resulting in low accuracy and high misdiagnosis rates in tic identification. Simultaneously, the lack of multi-dimensional anomaly detection and confidence quantification mechanisms makes it difficult to distinguish between mild tics and occasional movements. Summary of the Invention

[0003] This invention provides a digital method and system for monitoring psychological and behavioral tics, to solve the above-mentioned problems:

[0004] This invention proposes a digital method and system for monitoring psychological and behavioral tics, the method comprising:

[0005] S1. Using an AI chip, multiple frames of twitching monitoring images are acquired based on remote video information, a twitching monitoring coordinate system is established, initial static nodes are determined, and behavior collection information is obtained.

[0006] S2. Obtain the node dynamic behavior trajectory information of the initial static node based on the behavior acquisition information, delete the origin displacement information, and obtain the displacement behavior trajectory information.

[0007] S3. Perform differential anomaly analysis and soft clustering anomaly analysis on the initial static nodes using an AI chip, and then perform anomaly confidence analysis. Based on the obtained differential confidence analysis results and soft clustering confidence analysis results, obtain the corresponding abnormal nodes, perform troll confidence determination, and obtain troll confidence determination information.

[0008] Further, S1 includes:

[0009] The video stream parsing module of the AI ​​chip loads remote video information, and obtains multiple frames of twitching monitoring images based on the remote video information;

[0010] Acquire the initial twitching monitoring image from multiple frames of twitching monitoring images;

[0011] The AI ​​chip-based human key point detection model extracts human model information from the initial twitching monitoring image to obtain the initial human model information.

[0012] The coordinate calculation unit of the AI ​​chip establishes a rectangular coordinate system with the geometric center point of the initial human body model information as the origin, and obtains the twitching monitoring coordinate system.

[0013] Based on the initial human body model information, the AI ​​chip uses a node marking method to mark the key points in the motion monitoring coordinate system as initial static points, thereby obtaining initial static nodes.

[0014] The AI ​​chip's real-time data acquisition module obtains the displacement change information of the initial static node in the multi-frame twitching monitoring images, thereby acquiring behavioral acquisition information.

[0015] Further, S2 includes:

[0016] Based on the behavior acquisition information, the displacement change information of each initial static node is obtained, and then the node dynamic behavior trajectory information of the initial static node is obtained.

[0017] Synchronous displacement information of the dynamic behavior trajectory information of multiple initial static nodes is obtained through the synchronization analysis method of AI chip.

[0018] Obtain the origin displacement information of the geometric center point;

[0019] Determine whether the synchronous displacement information corresponds to the origin displacement information to obtain position correspondence determination information;

[0020] Based on the displacement corresponding judgment information, the synchronous displacement information is deleted from the node dynamic behavior trajectory information to obtain the initial static node displacement behavior trajectory information.

[0021] Further, S3 includes:

[0022] Obtain the preset behavior trajectory information of the initial static nodes;

[0023] The preset behavior trajectory information is compared with the displacement behavior trajectory information to obtain the difference information between the preset behavior trajectory information and the displacement behavior trajectory information;

[0024] The difference information is compared with a preset difference threshold to obtain the behavioral difference comparison result;

[0025] Based on the behavioral difference comparison results, determine the static nodes with differences;

[0026] The AI ​​chip calculates the difference confidence level of the static nodes based on the comparison information;

[0027] Based on the stated difference confidence level, a difference confidence analysis is performed to obtain the difference confidence analysis results.

[0028] Furthermore, based on the stated difference confidence level, a difference confidence analysis is performed to obtain the difference confidence analysis results, including:

[0029] The difference confidence level is compared with a preset difference confidence threshold;

[0030] When the difference confidence level is greater than the preset difference confidence threshold, the difference confidence level is determined to be qualified.

[0031] When the difference confidence level is less than or equal to the preset difference confidence threshold, the difference confidence level is deemed unqualified.

[0032] Furthermore, S3 also includes:

[0033] Set various soft clustering conditions;

[0034] Based on the aforementioned soft clustering condition information, the displacement behavior trajectory information of multiple initial static nodes is subjected to multiple iterative soft clustering analysis until the change in soft clustering information is less than a preset change threshold, thereby obtaining the physiological nodes corresponding to physiological displacement information and the pathological nodes corresponding to pathological displacement information.

[0035] The physiological confidence level of physiological nodes is calculated based on physiological displacement information using an AI chip.

[0036] The pathological confidence level of pathological nodes is calculated based on pathological displacement information using an AI chip.

[0037] The AI ​​chip performs soft clustering confidence analysis based on the physiological and pathological confidence levels to obtain the soft clustering confidence analysis results.

[0038] Furthermore, the AI ​​chip performs soft clustering confidence analysis based on the physiological confidence level and the pathological confidence level to obtain soft clustering confidence analysis results, including:

[0039] The physiological confidence level is compared with a preset physiological confidence threshold. When the physiological confidence level is greater than the preset physiological confidence threshold, the physiological confidence is deemed to be qualified.

[0040] The pathological confidence level is compared with a preset pathological confidence threshold. When the pathological confidence level is greater than the preset pathological confidence threshold, the pathological confidence is deemed to be qualified.

[0041] When both physiological and pathological confidence levels are met, soft clustering is considered to be of high confidence.

[0042] Furthermore, S3 also includes:

[0043] When the difference confidence level is qualified and the soft clustering confidence level is qualified, obtain the difference static node and pathological node;

[0044] By mapping the differential static nodes and pathological nodes, the corresponding abnormal nodes are obtained;

[0045] The abnormal twitching location information is determined based on the corresponding abnormal node;

[0046] Based on the data of the corresponding abnormal nodes, the confidence level of the twitching is determined, and the twitching confidence level determination information is obtained.

[0047] Further, the step of determining the confidence level of twitching based on the corresponding abnormal node data to obtain twitching confidence level determination information includes:

[0048] The number of abnormal node data is compared with a preset threshold for the number of abnormal nodes;

[0049] When the number of abnormal node data exceeds a preset threshold, the twitching confidence level is deemed acceptable.

[0050] When the number of abnormal node data is less than or equal to the preset node number threshold, the twitching confidence level is deemed unqualified.

[0051] Furthermore, the system includes:

[0052] The node determination module is used to acquire multiple frames of twitching monitoring images based on remote video information through an AI chip, establish a twitching monitoring coordinate system, determine the initial static nodes, and acquire behavior collection information.

[0053] The interference removal module is used to obtain the node dynamic behavior trajectory information of the initial static node based on the behavior acquisition information, delete the origin displacement information, and obtain the displacement behavior trajectory information.

[0054] The anomaly analysis module is used to perform differential anomaly analysis and soft clustering anomaly analysis on the initial static nodes through the AI ​​chip, and then perform anomaly confidence analysis. Based on the obtained differential confidence analysis results and soft clustering confidence analysis results, the corresponding abnormal nodes are obtained, and the troll confidence is determined to obtain troll confidence information.

[0055] The beneficial effects of this invention are: It enables tic behavior monitoring solely through remote video, avoiding the inconvenience of traditional face-to-face monitoring, while simultaneously achieving objective data analysis, improving data accuracy, and reducing subjectivity. Dual tic monitoring and analysis are achieved through difference analysis and soft clustering, reducing misdiagnosis rates and improving data analysis accuracy. A node number threshold quantifies the severity of tics. An origin displacement elimination algorithm eliminates interference from camera movement or overall body sway (such as global displacement during patient posture adjustments), ensuring the stability of local tic detection. A dual-confidence fusion mechanism (difference + soft clustering) avoids the limitations of single methods (e.g., soft clustering may misclassify nodes due to data noise, but difference analysis can provide secondary verification). High-confidence results are directly correlated with Tourette syndrome diagnostic data analysis. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of a digital psychological and behavioral tic monitoring method. Detailed Implementation

[0057] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0058] In one embodiment of the present invention, a digital psychological and behavioral tic monitoring method and system are proposed, the method comprising:

[0059] S1. Using an AI chip, multiple frames of twitching monitoring images are acquired based on remote video information, a twitching monitoring coordinate system is established, initial static nodes are determined, and behavior collection information is obtained.

[0060] S2. Obtain the node dynamic behavior trajectory information of the initial static node based on the behavior acquisition information, delete the origin displacement information, and obtain the displacement behavior trajectory information.

[0061] S3. Using an AI chip, perform differential anomaly analysis and soft clustering anomaly analysis on the initial static nodes, and then perform anomaly confidence analysis. Based on the obtained differential confidence analysis results and soft clustering confidence analysis results, obtain the corresponding abnormal nodes, perform troll confidence determination, and obtain troll confidence determination information, such as... Figure 1 As shown.

[0062] The working principle and technical effects of the above technical solution are as follows: A continuous frame image (e.g., 30 frames per second) is extracted from remote video using an AI chip, covering the entire cycle of the twitching motion. Based on the human skeleton key point detection method, a two-dimensional coordinate system is established in the image, and initial static nodes (e.g., shoulder joint, elbow joint, jaw, etc.) are marked to ensure node positioning accuracy. Node coordinates, timestamps, video frame numbers, and other information are recorded synchronously to form a structured behavior dataset. Based on the continuous frame coordinate data, the displacement change of each static node in the time series is calculated. The absolute displacement of each node relative to the reference point is calculated. The origin offset caused by the overall swaying of the human body is deleted, retaining the displacement related to local twitching (e.g., only the tremor of the shoulder relative to the torso). The displacement behavior trajectory is compared with a preset normal behavior template using Dynamic Time Warping (DTW), and the trajectory difference data is calculated. Difference nodes are marked based on the difference confidence threshold. Algorithms such as Fuzzy C-means (FCM) are used to perform soft clustering on the displacement trajectory, dividing it into physiological (normal) and pathological (twitching) clusters. High-probability pathological nodes are marked based on the soft clustering confidence threshold. Match the differential anomaly nodes with the soft clustering anomaly nodes. Count the number of successfully matched anomaly nodes. If the number is greater than a preset threshold, the spurious confidence level is considered acceptable; otherwise, it is considered unacceptable.

[0063] Remote video monitoring enables tic behavior monitoring, avoiding the inconvenience of traditional face-to-face monitoring, while simultaneously achieving objective data analysis, improving data accuracy, and reducing subjectivity. Dual tic monitoring analysis is achieved through differential analysis and soft clustering, reducing misdiagnosis rates and improving data analysis accuracy. A node number threshold quantifies the severity of tics. An origin displacement elimination algorithm eliminates interference from camera movement or overall body sway (such as global displacement during patient posture adjustments), ensuring the stability of localized tic detection. A dual-confidence fusion mechanism (differential analysis + soft clustering) avoids the limitations of single methods (e.g., soft clustering may misclassify nodes due to data noise, but differential analysis can provide secondary verification). High-confidence results are directly correlated with Tourette syndrome diagnostic data analysis.

[0064] In one embodiment of the present invention, S1 includes:

[0065] The video stream parsing module of the AI ​​chip loads remote video information, and obtains multiple frames of twitching monitoring images based on the remote video information;

[0066] Acquire the initial twitching monitoring image from multiple frames of twitching monitoring images;

[0067] The AI ​​chip-based human key point detection model extracts human model information from the initial twitching monitoring image to obtain the initial human model information.

[0068] The coordinate calculation unit of the AI ​​chip establishes a rectangular coordinate system with the geometric center point of the initial human body model information as the origin, and obtains the twitching monitoring coordinate system.

[0069] Based on the initial human body model information, the AI ​​chip uses a node marking method to mark the key points in the motion monitoring coordinate system as initial static points, thereby obtaining initial static nodes.

[0070] The AI ​​chip's real-time data acquisition module obtains the displacement change information of the initial static node in the multi-frame twitching monitoring images, thereby acquiring behavioral acquisition information.

[0071] The working principle and technical effects of the above technical solution are as follows: Multiple frames of images are decoded and filtered from a remote video stream, focusing on segments containing twitching movements. The first frame or a stable frame is selected as a reference, and initial human body model information is extracted using human keypoint detection methods (such as OpenPose). A Cartesian coordinate system is defined with the geometric center of the human body (such as the center of the pelvis or torso) as the origin, combined with the human body's orientation, ensuring that the coordinate system is relatively fixed relative to the human body. Nodes in the first frame (such as the mid-spine or hip joint) are selected as reference points, and their initial positions in the coordinate system are recorded. The positional changes of static nodes are continuously tracked across multiple frames, and parameters such as displacement and velocity relative to the initial coordinates are calculated to generate behavioral acquisition information (such as twitching frequency and amplitude).

[0072] By using human body modeling and coordinate system transformation, pixel-level motion is converted into physical spatial displacement (with millimeter-level precision), eliminating interference from perspective and distance. Static node design (such as the spine and hip joints) effectively distinguishes localized tics from overall body movement (such as walking and turning), reducing false positives. The coordinate system adaptively adjusts to the human body, avoiding noise introduced by camera shake or changes in patient position. It replaces subjective observation, providing quantitative evidence to support the diagnosis of Tourette syndrome. By comparing displacement characteristics before and after treatment (such as a 40% reduction in frequency), the effectiveness of medication or behavioral therapy can be quantified. It supports long-term data collection in a home environment, improving ecological effectiveness.

[0073] In one embodiment of the present invention, S2 includes:

[0074] Based on the behavior acquisition information, the displacement change information of each initial static node is obtained, and then the node dynamic behavior trajectory information of the initial static node is obtained.

[0075] Synchronous displacement information of the dynamic behavior trajectory information of multiple initial static nodes is obtained through the synchronization analysis method of AI chip.

[0076] Obtain the origin displacement information of the geometric center point;

[0077] Determine whether the synchronous displacement information corresponds to the origin displacement information to obtain position correspondence determination information;

[0078] Based on the displacement corresponding judgment information, the synchronous displacement information is deleted from the node dynamic behavior trajectory information to obtain the initial static node displacement behavior trajectory information.

[0079] The working principle and technical effect of the above technical solution are as follows: Based on behavior acquisition information (such as the coordinate changes of static nodes in multiple frames), the displacement time series of each initial static node is calculated to generate the node's dynamic behavior trajectory. Trajectory data from multiple static nodes are integrated to obtain their synchronous displacement information (such as the set of displacement vectors of all nodes at time t), reflecting the overall motion state of the human body. The displacement change of the geometric center point is monitored as a benchmark reference for the overall motion of the human body. It is determined whether the synchronous displacement information is consistent with the origin displacement information (such as a linear correlation between the displacement of all nodes and the origin displacement). Position correspondence judgment information is output (such as labels for "overall translation" or "local twitching"). If the synchronous displacement information corresponds to the origin displacement (i.e., the overall motion of the human body), that part of the displacement is deleted from the node's dynamic behavior trajectory (such as subtracting the origin displacement vector), retaining only the local displacement caused by twitching, thus obtaining pure displacement behavior trajectory information.

[0080] By correcting the origin displacement, interference from natural human movement (such as walking and turning) on ​​the static node trajectory is eliminated. This improves the specificity of tic detection (reducing false alarms caused by overall movement by more than 30%). It eliminates the need for external sensors (such as IMUs), distinguishing between camera perspective changes and genuine human tics solely through video analysis (e.g., slight camera shake does not affect the results). Synchronous displacement information integrates multi-node data, improving the robustness of motion state judgment (e.g., noise from a single node is compensated for by information from other nodes). It supports the recognition of complex tic patterns (e.g., simultaneously detecting coordinated tics involving shoulder shrugging and neck twisting). It provides clean tic trajectory data (e.g., neck twisting angle range ±15°), assisting doctors in developing personalized treatment plans. High-precision tic monitoring is achieved in a home environment, reducing the number of times patients need to visit the hospital (data acquisition efficiency improved by 50%).

[0081] In one embodiment of the present invention, S3 includes:

[0082] Obtain the preset behavior trajectory information of the initial static nodes;

[0083] The preset behavior trajectory information is compared with the displacement behavior trajectory information to obtain the difference information between the preset behavior trajectory information and the displacement behavior trajectory information;

[0084] The difference information is compared with a preset difference threshold to obtain the behavioral difference comparison result;

[0085] Based on the behavioral difference comparison results, determine the static nodes with differences;

[0086] The AI ​​chip calculates the difference confidence level of the static nodes based on the comparison information;

[0087] The formula for calculating the difference confidence level is as follows:

[0088]

[0089] Where CA is the difference confidence level, and T p For the preset behavioral trajectory information (vector), T o The observed displacement trajectory information (vector) is given, where ∆th is a preset difference threshold, and W... S W is the threshold weighting function; S (∆th) is the Sigmoid function, which nonlinearizes the effect of the threshold.

[0090] Based on the stated difference confidence level, a difference confidence analysis is performed to obtain the difference confidence analysis results.

[0091] Based on the aforementioned difference confidence level, a difference confidence analysis is performed to obtain the following results:

[0092] The difference confidence level is compared with a preset difference confidence threshold;

[0093] When the difference confidence level is greater than the preset difference confidence threshold, the difference confidence level is determined to be qualified.

[0094] When the difference confidence level is less than or equal to the preset difference confidence threshold, the difference confidence level is deemed unqualified.

[0095] The working principle and technical effect of the above technical solution are as follows: to obtain preset personnel behavior trajectory information, and then to obtain the actual monitored displacement behavior trajectory information of the movement according to the preset personnel behavior trajectory information;

[0096] The actual monitored displacement trajectory information is compared with the preset trajectory to quantify the difference between the two.

[0097] The difference information is compared with a preset difference threshold (such as a clinical experience value or a threshold output by a machine learning model) to generate behavioral difference comparison results, and the difference static nodes are marked according to the comparison results.

[0098] Calculate the difference confidence level (e.g., the probability value between 0 and 1).

[0099] If the confidence level is greater than the preset difference confidence threshold (e.g., 0.8), the difference confidence is considered acceptable. The difference analysis is relatively correct.

[0100] If the confidence level is less than or equal to the threshold, the result is deemed unqualified. The difference analysis may be inaccurate.

[0101] By dynamically comparing preset trajectory templates with actual data, atypical tic patterns can be detected, improving the accuracy of identifying complex tic types. Quantitative evaluation of difference confidence reduces subjective judgment bias. Confidence analysis filters out interference from random movements, and comprehensive evaluation through multiple parameters ensures the reliability of results. The preset trajectory library can be dynamically updated, supporting personalized monitoring schemes and reducing invalid data transmission.

[0102] In one embodiment of the present invention, S3 further includes:

[0103] Set various soft clustering conditions, including time, frequency, and frequency domain (displacement velocity).

[0104] Based on the aforementioned soft clustering condition information, the displacement behavior trajectory information of multiple initial static nodes is subjected to multiple iterative soft clustering analysis until the change in soft clustering information is less than a preset change threshold, thereby obtaining the physiological nodes corresponding to physiological displacement information and the pathological nodes corresponding to pathological displacement information.

[0105] The physiological confidence level of physiological nodes is calculated based on physiological displacement information using an AI chip.

[0106] The pathological confidence level of pathological nodes is calculated based on pathological displacement information using an AI chip.

[0107] The formula for calculating the physiological confidence level is as follows:

[0108]

[0109] Where SZ represents physiological confidence level, and T j For the displacement behavior trajectory of the current node, U s The average trajectory of the physiological cluster centers;

[0110] The formula for calculating the pathological confidence level is as follows:

[0111]

[0112] Where BZ represents the pathological confidence level, and U B Let be the average locus of the pathological cluster centers, and 'a' be the local minimum.

[0113] The AI ​​chip performs soft clustering confidence analysis based on the physiological and pathological confidence levels to obtain the soft clustering confidence analysis results.

[0114] The AI ​​chip performs soft clustering confidence analysis based on the physiological and pathological confidence levels to obtain soft clustering confidence analysis results, including:

[0115] The physiological confidence level is compared with a preset physiological confidence threshold. When the physiological confidence level is greater than the preset physiological confidence threshold, the physiological confidence is deemed to be qualified.

[0116] The pathological confidence level is compared with a preset pathological confidence threshold. When the pathological confidence level is greater than the preset pathological confidence threshold, the pathological confidence is deemed to be qualified.

[0117] When both physiological and pathological confidence levels are met, soft clustering is considered to be of high confidence.

[0118] The working principle and technical effects of the above technical solution are as follows:

[0119] Define a multidimensional feature space, including: time dimension, frequency dimension, frequency domain dimension (spectral characteristics of displacement velocity), etc.

[0120] Obtain the displacement behavior trajectory information (time series data) of multiple initial static nodes.

[0121] The trajectory data were soft-clustered using fuzzy C-means (FCM) or Gaussian mixture model (GMM), with each data point belonging to either the "physiological" or "pathological" cluster in terms of probability.

[0122] The cluster centers are iteratively optimized based on multidimensional conditional information (time, frequency, frequency domain) until the inter-cluster variation is less than a preset threshold (e.g., 0.01), ensuring convergence stability.

[0123] Physiological displacement information corresponds to physiological nodes, and pathological displacement information corresponds to pathological nodes.

[0124] If the physiological confidence level is greater than the preset physiological threshold (e.g., 0.85), the physiological confidence level is considered acceptable (e.g., confirming that nodding is a normal physiological action).

[0125] If the pathological confidence level is greater than the preset pathological threshold (e.g., 0.75), the pathological confidence level is deemed acceptable (e.g., confirming that neck twisting is a tics-related action).

[0126] Output "Soft clustering confidence is qualified" only when both physiological and pathological confidence levels are qualified; otherwise, mark it as data that needs to be reviewed (e.g., low confidence levels may be due to noise or mixed actions).

[0127] By combining time, frequency, and frequency domain information, the system addresses the problem of misclassification based on a single dimension, and multidimensional soft clustering improves the accuracy of physiological / pathological classification. It allows data points to belong to both physiological and pathological clusters simultaneously, quantifying uncertainty through probability values ​​to avoid the "either / or" error of hard clustering. Preset thresholds can be dynamically adjusted based on clinical data, supporting personalized monitoring plans.

[0128] High-confidence pathological nodes are directly associated with Tourette syndrome diagnoses. Changes in pathological confidence before and after treatment quantify the effectiveness of medication or behavioral therapy. Invalid data transmission is reduced (only pathological trajectories with acceptable confidence are uploaded), thus lowering bandwidth requirements.

[0129] In one embodiment of the present invention, S3 further includes:

[0130] When the difference confidence level is qualified and the soft clustering confidence level is qualified, obtain the difference static node and pathological node;

[0131] By mapping the differential static nodes and pathological nodes, the corresponding abnormal nodes are obtained;

[0132] The abnormal twitching location information is determined based on the corresponding abnormal node;

[0133] Based on the data of the corresponding abnormal nodes, the confidence level of the twitching is determined, and the twitching confidence level determination information is obtained.

[0134] The step of determining the confidence level of twitching based on the corresponding abnormal node data to obtain twitching confidence level information includes:

[0135] The number of abnormal node data is compared with a preset threshold for the number of abnormal nodes;

[0136] When the number of abnormal node data exceeds a preset threshold, the twitching confidence level is deemed acceptable.

[0137] When the number of abnormal node data is less than or equal to the preset node number threshold, the twitching confidence level is deemed unqualified.

[0138] The working principle and technical effect of the above technical solution are as follows: the difference confidence qualified means that the judgment that there is a significant difference between the actual displacement trajectory and the preset behavior trajectory is accurate.

[0139] A qualified soft clustering confidence level indicates that the soft clustering analysis accurately distinguishes between physiological and pathological nodes.

[0140] Obtain differential static nodes from differential confidence-qualified results, and obtain pathological nodes from soft clustering confidence-qualified results.

[0141] The differential static nodes are matched with pathological nodes, and the specific abnormal sites are marked based on the matching results.

[0142] Calculate the total number of corresponding abnormal nodes.

[0143] If the number of abnormal nodes is greater than the preset threshold (e.g., 2), the confidence level of the tics is deemed acceptable (e.g., significant tics are confirmed).

[0144] If the number of abnormal nodes is less than or equal to the threshold, the result is deemed unqualified (e.g., it may be due to accidental action or noise interference).

[0145] Differential confidence scores filter out random movements, ensuring the significance of trajectory deviations. Soft clustering confidence scores distinguish between physiological and pathological movements, avoiding misdiagnosis. Node number thresholds quantify tic analysis results. Node matching is only performed when both confidence scores are satisfactory, reducing the limitations of single analysis methods. High-confidence results are directly correlated with Tourette syndrome diagnosis.

[0146] Treatment data is quantified by measuring changes in the number of abnormal nodes before and after treatment. Only tic events with acceptable confidence levels are uploaded to reduce invalid data transmission.

[0147] The preset thresholds can be dynamically adjusted based on clinical data (such as the difference in abnormal node thresholds between children and adults), supporting personalized monitoring programs.

[0148] In one embodiment of the present invention, the system includes:

[0149] The node determination module is used to acquire multiple frames of twitching monitoring images based on remote video information through an AI chip, establish a twitching monitoring coordinate system, determine the initial static nodes, and acquire behavior collection information.

[0150] The interference removal module is used to obtain the node dynamic behavior trajectory information of the initial static node based on the behavior acquisition information, delete the origin displacement information, and obtain the displacement behavior trajectory information.

[0151] The anomaly analysis module is used to perform differential anomaly analysis and soft clustering anomaly analysis on the initial static nodes through the AI ​​chip, and then perform anomaly confidence analysis. Based on the obtained differential confidence analysis results and soft clustering confidence analysis results, the corresponding abnormal nodes are obtained, and the troll confidence is determined to obtain troll confidence information.

[0152] The working principle and technical effect of the above technical solution are as follows: extract continuous frame images (e.g., 30 frames per second) from remote video to cover the entire cycle of the twitching motion. Based on the human skeleton key point detection method, a two-dimensional coordinate system is established in the image, and initial static nodes (such as shoulder joint, elbow joint, jaw, etc.) are marked to ensure node positioning accuracy.

[0153] Synchronously record node coordinates, timestamps, video frame numbers, and other information to form a structured behavioral dataset.

[0154] Based on continuous frame coordinate data, calculate the displacement change of each static node over time.

[0155] Calculate the absolute displacement of each node relative to the reference point.

[0156] Remove the origin offset caused by the overall swaying of the human body, and retain the displacement related to local twitches (such as only retaining the swaying of the shoulders relative to the torso).

[0157] The displacement behavior trajectory is compared with the preset normal behavior template by dynamic time warping (DTW) to calculate the trajectory difference data.

[0158] By combining the difference confidence threshold, the difference nodes are marked.

[0159] Algorithms such as Fuzzy C-means (FCM) were used to perform soft clustering of displacement trajectories, dividing them into physiological (normal) and pathological (tic) clusters.

[0160] By combining the confidence threshold of soft clustering, high-probability pathological nodes are labeled.

[0161] Match the differential anomaly nodes with the soft clustering anomaly nodes.

[0162] The number of successfully matched abnormal nodes is counted. If the number is greater than the preset threshold, the tug confidence level is considered qualified; otherwise, it is considered unqualified.

[0163] Tic behavior monitoring can be achieved through remote video, avoiding the inconvenience of traditional face-to-face monitoring, while enabling objective data analysis, improving data accuracy, and reducing subjectivity.

[0164] By employing both differential analysis and soft clustering analysis, dual tic-twitch monitoring and analysis can be achieved, reducing the misdiagnosis rate and improving the accuracy of data analysis.

[0165] The severity of the disturbance is quantified by the number of nodes.

[0166] The origin displacement elimination algorithm eliminates interference from camera movement or overall human body sway (such as global displacement when the patient adjusts their sitting posture), ensuring the stability of local twitching detection.

[0167] The dual confidence fusion mechanism (difference + soft clustering) avoids the limitations of a single method (such as soft clustering may misclassify nodes due to data noise, but difference analysis can perform secondary verification).

[0168] High-confidence results are directly correlated with the analysis of Tourette syndrome diagnostic data.

[0169] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A digital method for monitoring psychological and behavioral tic disorders, characterized in that, The method includes: S1. Using an AI chip, multiple frames of twitching monitoring images are acquired based on remote video information, a twitching monitoring coordinate system is established, initial static nodes are determined, and behavior collection information is obtained. S2. Obtain the node dynamic behavior trajectory information of the initial static node based on the behavior acquisition information, delete the origin displacement information, and obtain the displacement behavior trajectory information. S3. Perform differential anomaly analysis and soft clustering anomaly analysis on the initial static nodes using an AI chip, and then perform anomaly confidence analysis. Based on the obtained differential confidence analysis results and soft clustering confidence analysis results, obtain the corresponding abnormal nodes, perform troll confidence determination, and obtain troll confidence determination information. Wherein, S2 includes: Based on the behavior acquisition information, the displacement change information of each initial static node is obtained, and then the node dynamic behavior trajectory information of the initial static node is obtained. Synchronous displacement information of the dynamic behavior trajectory information of multiple initial static nodes is obtained through the synchronization analysis method of AI chip. Obtain the origin displacement information of the geometric center point; Determine whether the synchronous displacement information corresponds to the origin displacement information to obtain position correspondence determination information; Based on the position correspondence judgment information, the synchronous displacement information is deleted from the node dynamic behavior trajectory information to obtain the displacement behavior trajectory information of the initial static node; Wherein, S3 includes: Obtain the preset behavior trajectory information of the initial static nodes; The preset behavior trajectory information is compared with the displacement behavior trajectory information to obtain the difference information between the preset behavior trajectory information and the displacement behavior trajectory information; The difference information is compared with a preset difference threshold to obtain the behavioral difference comparison result; Based on the behavioral difference comparison results, determine the static nodes with differences; The AI ​​chip calculates the difference confidence level of the static nodes based on the comparison information; Based on the stated difference confidence level, a difference confidence level analysis is performed to obtain the difference confidence level analysis results; S3 further includes: Set various soft clustering conditions; Based on the aforementioned soft clustering condition information, the displacement behavior trajectory information of multiple initial static nodes is subjected to multiple iterative soft clustering analysis until the change in soft clustering information is less than a preset change threshold, thereby obtaining the physiological nodes corresponding to physiological displacement information and the pathological nodes corresponding to pathological displacement information. The physiological confidence level of physiological nodes is calculated based on physiological displacement information using an AI chip. The pathological confidence level of pathological nodes is calculated based on pathological displacement information using an AI chip. The AI ​​chip performs soft clustering confidence analysis based on the physiological and pathological confidence levels to obtain the soft clustering confidence analysis results.

2. The digital psychological and behavioral tic monitoring method according to claim 1, characterized in that, S1 includes: The video stream parsing module of the AI ​​chip loads remote video information, and obtains multiple frames of twitching monitoring images based on the remote video information; Acquire the initial twitching monitoring image from multiple frames of twitching monitoring images; The AI ​​chip-based human key point detection model extracts human model information from the initial twitching monitoring image to obtain the initial human model information. The coordinate calculation unit of the AI ​​chip establishes a rectangular coordinate system with the geometric center point of the initial human body model information as the origin, and obtains the twitching monitoring coordinate system. Based on the initial human body model information, the AI ​​chip uses a node marking method to mark the key points in the motion monitoring coordinate system as initial static points, thereby obtaining initial static nodes. The AI ​​chip's real-time data acquisition module obtains the displacement change information of the initial static node in the multi-frame twitching monitoring images, thereby acquiring behavioral acquisition information.

3. The digital psychological and behavioral tic monitoring method according to claim 1, characterized in that, Based on the aforementioned difference confidence level, a difference confidence analysis is performed to obtain the following results: The difference confidence level is compared with a preset difference confidence threshold; When the difference confidence level is greater than the preset difference confidence threshold, the difference confidence level is determined to be qualified. When the difference confidence level is less than or equal to the preset difference confidence threshold, the difference confidence level is deemed unqualified.

4. The digital psychological and behavioral tic monitoring method according to claim 1, characterized in that, The AI ​​chip performs soft clustering confidence analysis based on the physiological and pathological confidence levels to obtain soft clustering confidence analysis results, including: The physiological confidence level is compared with a preset physiological confidence threshold. When the physiological confidence level is greater than the preset physiological confidence threshold, the physiological confidence is deemed to be qualified. The pathological confidence level is compared with a preset pathological confidence threshold. When the pathological confidence level is greater than the preset pathological confidence threshold, the pathological confidence is deemed to be qualified. When both physiological and pathological confidence levels are met, soft clustering is considered to be of high confidence.

5. The digital psychological and behavioral tic monitoring method according to claim 1, characterized in that, S3 further includes: When the difference confidence level is qualified and the soft clustering confidence level is qualified, obtain the difference static node and pathological node; By mapping the differential static nodes and pathological nodes, the corresponding abnormal nodes are obtained; The abnormal twitching location information is determined based on the corresponding abnormal node; Based on the data of the corresponding abnormal nodes, the confidence level of the twitching is determined, and the twitching confidence level determination information is obtained.

6. The digital psychological and behavioral tic monitoring method according to claim 5, characterized in that, The step of determining the confidence level of twitching based on the corresponding abnormal node data to obtain twitching confidence level information includes: The number of abnormal node data is compared with a preset threshold for the number of abnormal nodes; When the number of abnormal node data exceeds a preset abnormal node number threshold, the tug confidence level is deemed qualified. When the number of abnormal node data is less than or equal to a preset abnormal node number threshold, the twitching confidence level is deemed unqualified.

7. A digital psychological and behavioral tic monitoring system, characterized in that, The system includes: The node determination module is used to acquire multiple frames of twitching monitoring images based on remote video information through an AI chip, establish a twitching monitoring coordinate system, determine the initial static nodes, and acquire behavior collection information. The interference removal module is used to obtain the node dynamic behavior trajectory information of the initial static node based on the behavior acquisition information, delete the origin displacement information, and obtain the displacement behavior trajectory information. The anomaly analysis module is used to perform differential anomaly analysis and soft clustering anomaly analysis on the initial static nodes through the AI ​​chip, and then perform anomaly confidence analysis. Based on the obtained differential confidence analysis results and soft clustering confidence analysis results, the corresponding abnormal nodes are obtained, and the troll confidence is determined to obtain troll confidence determination information. The interference removal module includes: Based on the behavior acquisition information, the displacement change information of each initial static node is obtained, and then the node dynamic behavior trajectory information of the initial static node is obtained. Synchronous displacement information of the dynamic behavior trajectory information of multiple initial static nodes is obtained through the synchronization analysis method of AI chip. Obtain the origin displacement information of the geometric center point; Determine whether the synchronous displacement information corresponds to the origin displacement information to obtain position correspondence determination information; Based on the position correspondence judgment information, the synchronous displacement information is deleted from the node dynamic behavior trajectory information to obtain the displacement behavior trajectory information of the initial static node; The anomaly analysis module includes: Obtain the preset behavior trajectory information of the initial static nodes; The preset behavior trajectory information is compared with the displacement behavior trajectory information to obtain the difference information between the preset behavior trajectory information and the displacement behavior trajectory information; The difference information is compared with a preset difference threshold to obtain the behavioral difference comparison result; Based on the behavioral difference comparison results, determine the static nodes with differences; The AI ​​chip calculates the difference confidence level of the static nodes based on the comparison information; Based on the stated difference confidence level, a difference confidence level analysis is performed to obtain the difference confidence level analysis results; The anomaly analysis module further includes: Set various soft clustering conditions; Based on the aforementioned soft clustering condition information, the displacement behavior trajectory information of multiple initial static nodes is subjected to multiple iterative soft clustering analysis until the change in soft clustering information is less than a preset change threshold, thereby obtaining the physiological nodes corresponding to physiological displacement information and the pathological nodes corresponding to pathological displacement information. The physiological confidence level of physiological nodes is calculated based on physiological displacement information using an AI chip. The pathological confidence level of pathological nodes is calculated based on pathological displacement information using an AI chip. The AI ​​chip performs soft clustering confidence analysis based on the physiological and pathological confidence levels to obtain the soft clustering confidence analysis results.

Images